Utilising family-based designs for detecting rare variant disease associations

Abstract

Rare genetic variants are thought to be important components in the causality of many diseases but discovering these associations is challenging. We demonstrate how best to use family-based designs to improve the power to detect rare variant disease associations. We show that using genetic data from enriched families (those pedigrees with greater than one affected member) increases the power and sensitivity of existing case-control rare variant tests. However, we show that transmission- (or within-family-) based tests do not benefit from this enrichment. This means that, in studies where a limited amount of genotyping is available, choosing a single case from each of many pedigrees has greater power than selecting multiple cases from fewer pedigrees. Finally, we show how a pseudo-case-control design allows a greater range of statistical tests to be applied to family data.

Keywords: Rare variants; TDT; family-based association; missing heritability.

Figure 1

Results for three simulated scenarios. The power is presented in each case with the results for trio derived data on the left, ASP derived data in the middle and enriched trios on the right. (A) The number of causal variants increases from 1 to 10 and the OR is varied to maintain an approximate 90% highest absolute power. (B) For four causal variants we jointly vary their ORs from 1 to 2. (C) While keeping four causal variants we change the direction of the effect of an increasing number of them, from four causal and no protective variants (the baseline scenatio, left) to four protective and no casual variants (right).

Figure 2

We show the relative power of each of the five score tests to the power of the score tests from data derived from 600 ASPs (1200 cases, black). Results are presented from data derived from 300 ASPs (600 cases, blue), 600 enriched trios (600 cases, green) and 900 trios (900 cases, red). The family-based tests are on the left, the pseudo-case–controls in the middle and unrelated-case–controls on the right.